Sr. Fullstack Platform Engineer - (Backend Focus)

Salt Ai•Los Angeles, CA

9h•Remote

About The Position

Salt AI is building the governed execution platform for AI work that has to be real: deployable, auditable, permission-aware, and reliable enough for regulated industries. Our customers are doing work where “the demo looked cool” is not enough. In life sciences, that can mean RNA-targeted drug discovery, combinatorial chemistry, clinical-trial-adjacent research, and scientific workflows that need traceability, reproducibility, and trust. In financial services, healthcare, legal, and government, it means sensitive data, private infrastructure, and AI systems that have to be controlled rather than merely impressive. We are looking for a Senior Fullstack Platform Engineer who can help turn that platform into something customers can actually build on: software that feels considered, powerful, fast, trustworthy, and unusually good for the complexity underneath. This role leans backend and platform. You should go deep on Python, Django, APIs, cloud infrastructure, Kubernetes, workflow execution, data systems, reliability, and platform contracts. But the scope is still fullstack. You are not just building backend services in isolation. You are responsible for the Fullstack of the customer experience: infrastructure, data contracts, APIs, runtime behavior, UI surfaces, failure states, and the quality bar of the thing a customer actually uses.

Requirements

Strong backend and platform experience.
Credible in Python/Django, APIs, distributed systems, cloud infrastructure, Kubernetes, queues, jobs, and production reliability.
Comfortable enough with TypeScript, React, and frontend product surfaces to build, debug, or shape the UI that exposes your backend work.
High agency and high standards. You do not wait for a perfect spec. You clarify the problem, find the constraint, make progress, and raise the quality bar as you go.
Platform instincts. You think in contracts, interfaces, failure modes, permissions, observability, and lifecycle. You know the difference between a feature that works once and a primitive that other people can safely build on.
Good taste in abstraction. You do not over-framework the first version, but you can see when repeated customer work wants to become a platform capability.
Taste in customer experience. You can tell when a complex workflow is technically correct but still confusing, brittle, or hard to trust. You care about making advanced software feel legible, fast, and empowering.
AI-first engineering habits. You use tools like Claude Code, Cursor, Codex, or similar systems to move faster and think at a higher level. You still understand the code you ship, review generated work carefully, and know when to slow down.
Judgment in spite of AI. You can deliver high-quality software even when AI tools are eager to generate too much code, plausible abstractions, brittle tests, or shallow solutions.
Customer empathy. You can talk to scientists, data teams, operators, and enterprise stakeholders, then translate messy real-world needs into durable engineering decisions.
Clear communication. You can write down the shape of a problem, explain tradeoffs, and help the team make better decisions without turning everything into a meeting.

Nice To Haves

You have built workflow systems, developer platforms, data platforms, infrastructure products, agent systems, internal tools, or complex enterprise SaaS used by technical customers.
You have opinions about API shape, execution semantics, logs, permissions, lifecycle states, and observability because you have seen what happens when those things are treated as afterthoughts.
You can show how AI has made you faster without making your work worse.
You can point to places where you rejected, rewrote, constrained, or heavily edited generated code because your judgment was better than the tool’s first answer.
You have pulled something back from release because it technically worked but did not yet meet the quality bar.
You are comfortable with ambiguity, but you do not confuse ambiguity with vagueness. You ask the questions that make the work concrete.
You care about regulated, high-trust AI because it is harder and more useful than another thin wrapper around a chat box.

Responsibilities

Build the contracts that let agents call real pipelines with typed inputs, structured outputs, streaming status, retries, logs, and auditability.
Model platform concepts cleanly, build durable APIs, evolve service boundaries, and make customer workflows reliable at the backend layer.
Work on deployment patterns, scaling, isolation, runtime behavior, observability, and operational reliability for AI workflows.
Build the frontend and backend experiences customers use to configure, run, inspect, debug, and reuse AI workflows.
Build secure customer data access patterns, retrieval systems, indexing, hybrid search, and APIs that respect enterprise boundaries.
Make distributed AI workflows understandable: what ran, what changed, what failed, why it failed, and what the user or system can do next.
Turn one-off customer implementations into reusable capabilities without sanding off the important domain-specific details.
Improve the tools, docs, test harnesses, internal workflows, and diagnostics that let a small team move quickly without losing control of the system.